On Oct 12, 2007, at 2:19 AM, Peter Teoh wrote:
On 9/27/07, Bruce Rowen <browen@xxxxxxxxxxxx> wrote:
[snip]
Here is the strange part:
According to my internal counter and the jiffies counter. The
'dropped' tasklets occur on _exactly_ 10 minute boundaries (600,000
jiffies). Never one less or one more, always 600k.
I tried running my tasklets with tasklet_hi_schedule() which should
put them above timers and ethernet but observe no changes. Whatever
is delaying my tasklets is blocking about 8 milliseconds of cpu every
Please elaborate on "dropped tasklets"? Do u mean that the scheduler
did not execute it at the interval you specified? If so, then I
think this (full kernel preemptibility) is not guaranteed by Linux
kernel. A lot of complex scenario (which I cannot cite or imagine)
may happen between having 256M vs 1G ram, which result in kernel
preemption not as successful as in the 256M version. You may try out
the -rt (realtime) branch of the kernel. But if you mean that the
tasks just disappeared without the tasklet itself exiting, then I
think that will be a bug in the kernel.
Thanks for the reply Peter.
The tasklets are never run. I have a counter that increments by the
ISR (very short) and all the tasklets increment their own counters
wheb they run. I have a watchdog that barks when the tasklets detect
that the ISR counter has gained a tick over the tasklet counter.
Yes it is puzzling, seems to be memory size related and the
corresponding jiffy count is exactly 600000 higher than the last time
the hiccup occurs. Some kernel process is firing off exactly every
ten minutes.
-Bruce
Sorry, cannot help much, may be others can comment?
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
